Alibaba Group Holding Limited

Alibaba Group Holding Limited

81.08
USD
-2.23
(-2.68%)
7.00EPS
14.42P/E
267.53BMarket Cap
Mar 25Next Earn

Q2 FY2027 · Earnings Call TranscriptNovember 19, 2026

Operator

Good day, ladies and gentlemen. Thank you for standing by.

Welcome to Alibaba Group's September Quarter Earnings Conference Call. At this time, all participants are in listen-only mode.

After management's prepared remarks, there will be a Q&A session. I would now like to hand the call over to Lydia Liu, Head of Investor Relations of Alibaba Group.

Please go ahead.

Lydia Liu

Good day, everyone, and welcome to Alibaba Group's September Quarter 2026 Earnings Conference Call. Joining me today are Joe Tsai, Chairman; Eddie Wu, Chief Executive Officer; Toby Xu, Chief Financial Officer; Jiang Fan, Chief Executive Officer of Alibaba E-commerce Business Group.

I will quickly cover the safe harbor. Our remarks today will include forward-looking statements based on our current expectations, which are subject to risks and uncertainties.

We will discuss certain non-GAAP financial measures. Reconciliations between GAAP and non-GAAP measures are included in today's earnings press release and investor presentation.

Our comments will be on year-over-year comparisons unless otherwise stated. A replay of the call will be available on our website later today.

I will turn the call over to Eddie. 各位投资者朋友,欢迎参加。

Greetings and welcome to Alibaba Group's earnings call for the second quarter of fiscal year 2027. This quarter, Alibaba Group revenue growth accelerated, and profitability continued to improve.

Group total revenue grew 10% year-over-year. In AI cloud and compute services, external revenue growth accelerated to 64%.

At the same time, the year-over-year growth rate of recurring revenue from external customers increased to 67%. Adjusted EBITA margin rose to 14.5%.

AI-related product revenue grew at a triple-digit rate for the 13th consecutive quarter, and its annual revenue run rate exceeded $10 billion. Profitability improved significantly across our e-commerce businesses, and in quick commerce, we continued to narrow losses substantially while maintaining market share.

The rapid improvement of AI model capability is accelerating the adoption of AI applications across industries and driving demand for training and inference. The deployment of AI agents at scale is further driving demand for compute.

The combined result of these factors is that global compute demand will outstrip supply over the next several years. This profound demand is providing a solid foundation for AI commercialization at scale.

Next, I'll discuss four areas in more detail: AI and cloud commercialization, T-Head AI chips, models and AI applications, and our consumer business. First, both revenue and profit growth in AI and cloud are accelerating.

This quarter, revenue from external customers in AI cloud and compute services grew 64% year-over-year. That was the 10th consecutive quarter of accelerating growth.

Adjusted EBITA increased 121% year-over-year, and EBITA margin expanded to 14.5%. This performance was driven by growth across compute, storage, MaaS, and AI applications and synergies.

AI-related products generated RMB17.9 billion in quarterly revenue and accounted for 49% of Alibaba Cloud's external recurring revenue. AI is now a core revenue pillar of our cloud business.

The annual revenue run rate for AI-related products exceeded $10 billion. The strong revenue growth reflects broad demand for our AI-plus-cloud services.

Beyond demand from AI labs for training and inference, we continue to see strong demand from customers across various industries in China, including large state-owned enterprises as well as companies in intelligent manufacturing, automotive, embodied AI, and biotechnology. Second, we continue to scale up T-Head AI chip volumes and strengthen our supply chain capabilities.

Starting with chips, T-Head has established a portfolio of proprietary chips spanning the full stack of GPU, CPU, storage, and networking chips. Since launching in August, Supernode instances powered by Zhenwu M890 have ramped up in volume and are now being commercialized at scale.

We expect to begin mass production of the next-generation Zhenwu V900 chip in the first quarter of calendar 2027. The Zhenwu V900 delivers three times the computing performance of the M890.

Next-generation Supernodes powered by the Zhenwu V900 support large-scale network interconnections, enabling workloads such as training and inference for ultra-large-scale models. These Supernodes will be the cornerstone of Alibaba Cloud's AI infrastructure.

Zhenwu V900 Supernode clusters will launch at commercial scale in the second quarter of 2027. T-Head is accelerating its three-generation GPU road map.

We fully rolled out Zhenwu M890 this year. We plan to ramp up the supply chain and move into large-scale production for the V900 in 2027 and the J900 in 2028.

Infrastructure powered by our own domestically developed chips is becoming the foundation of Alibaba's full-stack AI capabilities. It will increasingly support model training, inference, and agentic workloads for both Alibaba Group and Alibaba Cloud customers, supporting growth in the group's AI revenue.

Qwen is already training its next-generation model on T-Head's domestic chips, and Model Studio is already providing model inference on domestic chips. Our full-stack AI infrastructure will give us a clear advantage in the domestic compute supply chain.

Deep integration between our proprietary chips and models will continue to lower the cost per unit of compute and help us gain greater efficiency in AI application development. Taking all of these factors into account, our targets for calendar year 2030 remain unchanged: achieving $100 billion in external cloud revenue with a margin exceeding 20%.

Third, we continue to advance our models and AI agent applications. Our newly released foundation model, Qwen 4-Max, has 2.4 trillion backbone parameters.

We continue to advance its key capabilities, including agentic tasks and coding, where it's among the leading models globally. Top-tier partners have also recognized Qwen's capabilities.

Apple Intelligence, which is deeply integrated with Qwen, is set to launch for users in China in December. On the enterprise side, Qwen Work launched in August and quickly became a leading enterprise AI agent.

We recently launched Qwen Work E-commerce, and the percentage of paying e-commerce users continues to rise. On the consumer side, the Qwen app is becoming a personalized assistant for getting things done.

It now offers services for more than 20 daily-life use cases. Fourth, e-commerce businesses' profitability grew significantly with improved China e-commerce EBITA and margin.

This quarter, our e-commerce businesses delivered steady revenue growth and significantly higher profits. China e-commerce adjusted EBITA and margins also significantly improved, demonstrating its stable profitability.

Our quick commerce business maintained its market share and continued to narrow losses, with unit economics improving quarter-over-quarter. Two quarters ago, we began to see our AI investments cross the commercialization inflection point.

Last quarter, growth accelerated, margins expanded. This quarter, those investments are delivering commercial results faster, and the returns are clearer.

Our multi-year investment in compute and cloud infrastructure has given us full-stack product advantages and economies of scale, which are becoming a key competitive advantage for Alibaba in the AI industry. They allow us to respond flexibly to industry shifts and offer AI products at scale to meet our customers' rising demands for AI.

Looking ahead, we will continue to invest in AI infrastructure to extend our lead and strengthen our long-term competitiveness. I will now hand over to Toby for the financial update.

Thank you.

Toby Xu

Thank you, Eddie. This quarter, our business delivered strong growth in both revenue and earnings because of our strategic focus on AI-plus-cloud growth and profit sustainability in consumption business.

Revenue from AI cloud and compute services grew 53% year-over-year, with EBITA margin sequentially rising to 14.5%. The growth of recurring revenue from external customers accelerated to 67% year-over-year.

The strong growth reflected broader-based demand for our AI-plus-cloud services. On consumption, EBITA of our e-commerce group achieved 243% year-over-year growth this quarter, primarily driven by a significant loss reduction in quick commerce.

In addition, our China e-commerce business contributed to stable operating profit, and the global wholesale business delivered strong profits while we reduced the losses in the international e-commerce business. Now let's look at the financial results for this quarter.

Total revenue increased 10% year-over-year to RMB272 billion, driven by the strong momentum in cloud business and quick commerce. Total adjusted EBITA increased 260% to RMB32.6 billion, primarily attributable to the improved operating results in Taobao Instant Commerce and cloud business, as well as enhanced operating efficiencies across various businesses.

GAAP net income was RMB64.4 billion, an increase of 213%, primarily driven by the increase in net investment income and income from operations. Operating cash flow was strong this quarter, increasing 55% year-over-year to RMB15.7 billion.

Capex was RMB50.5 billion, or $7.5 billion, this quarter due to continued investments in AI infrastructure to meet strong and accelerating customer demand. The significant year-over-year increase was due to an acceleration in scaling compute capacity, higher pricing for a broader range of chip components, and increasing CPU compute capacity driven by anticipated growing customer adoption of AI agents.

The quarter-over-quarter decline from RMB67.7 billion, or $10 billion, in the previous quarter reflects fluctuations in procurement cycles. We will maintain our disciplined approach by consistently tracking and evaluating the projected ROIC for future capex.

Free cash flow was an outflow of RMB34.5 billion compared to an outflow of RMB21.8 billion in the same quarter last year. The decrease was mainly attributable to the investment in cloud infrastructure, partly offset by reduced losses in Taobao Instant Commerce and increase in customer prepayments for our cloud services.

The quarter-over-quarter improvement from an outflow of RMB44.7 billion last quarter reflects our prudent cash management. As of September 30, 2026, we continued to maintain a strong net cash position of $32.9 billion, up from $30.7 billion in the previous quarter.

Excluding debt with maturities beyond five years, our net cash position was at approximately $49 billion. Our robust operating cash flow and net cash position, together with realized investment gains, give us the confidence and the financial flexibility to invest in long-term growth.

We will maintain disciplined investments in AI capabilities with a focus on return on invested capital while targeting to achieve positive free cash flow. Let me now turn to our segment results.

Alibaba E-commerce Group: we sustained revenue stability with CMR decreasing 5% year-over-year and growing 1% on a like-for-like basis if the contra-revenue impact from the new business development program were excluded. Revenue from China quick commerce business was RMB58.8 billion, an increase of 22%, primarily driven by Freshippo's robust growth across online and offline channels.

Alibaba E-commerce Group's adjusted EBITA grew 243% to RMB38.9 billion, primarily driven by a significant loss reduction in the China quick commerce business. Taobao Instant Commerce continued to improve its UE, reflecting higher average order value.

Marketplace e-commerce delivered strong earnings supported by stable operating profit of the China e-commerce business, a loss reduction in the international e-commerce business, and strong growth in the global wholesale business. Now let's review the business updates and results of AI cloud and compute services.

Total revenue was RMB61 billion, a 53% increase from the same quarter last year. Revenue from external customers was RMB47 billion, with growth accelerating to 64%.

Recurring revenue from external customers, which excludes compute equipment solutions revenue, grew 67% to RMB37 billion. Customers continued to expand their use of our AI-plus-cloud services, driven by strong demand for AI computing and core CPU services, including storage, compute, security, and networking.

This quarter's AI-related product revenue was RMB17.9 billion, implying an annual revenue run rate of RMB71. 7 billion.

It delivered the 13th consecutive quarter of triple-digit growth and accounted for 49% of external recurring revenue. The adjusted EBITA margin expanded to 14.9%, benefiting from pricing enhancements amid tight market supply and scale-driven operating leverage.

We reiterate our calendar year 2030 guidance of achieving $100 billion in external cloud revenue with a long-term margin exceeding 20%, driven by broader enterprise adoption and deeper integration of our AI products into core business workflows. Next, let's look at AI Labs and Applications.

Total revenue was RMB4.4 billion, a 51% increase from the same quarter last year, primarily driven by rising demand for agentic AI products. The segment narrowed loss quarter-over-quarter to RMB10.5 billion.

We observed improved efficiency in both model training and marketing spend on the Qwen app. We are also continuously upgrading our AI agent offerings for both businesses and consumers through innovations across Qwen Work and the Qwen app.

In October, we introduced Qwen Work E-commerce, which is integrated into Qwen Work and is able to automate end-to-end cross-platform operations for e-commerce merchants. Additionally, we have launched QwenGo, a cloud-based personal agent embedded within the Qwen app.

All other segment revenue decreased 5% year-over-year to RMB27.7 billion, reflecting the decrease in revenue from Cainiao. All other adjusted EBITA was a loss of RMB3.3 billion, primarily due to our increased investment in technology businesses.

In conclusion, these strong financial results demonstrate that advances across our co-designed, integrated AI stack, together with the profit sustainability of our consumption business, position us well to capture increasing diverse demand across industries and create greater value for our shareholders. Building on this momentum, we will continue to balance disciplined investments for growth with a focus on return on invested capital and proactive cash management to accelerate our journey to positive free cash flow.

Thank you. That's the end of our prepared remarks.

We can open up for Q&A.

Lydia Liu

Thank you, Toby. We will now begin the Q&A session.

You are welcome to ask questions in Chinese or English. A third-party translator will provide consecutive interpretation.

In the case of any discrepancy, our management statements in the original language will prevail. Operator, let's go with the first question.

Thank you.

Operator

Thank you. Your first question comes from Gary Yu with Morgan Stanley.

Gary Yu

Hi, good evening. Thank you, management, for taking my questions.

My question is about the cloud revenue outlook. I appreciate the guidance of $100 billion in 2030, but in light of the latest BIS announcement, how should we look at the cloud revenue growth in the next two quarters within the fiscal year?

Are we still confident to deliver over 50% revenue growth going into next fiscal year in fiscal 2028? If so, what makes up confidence about it in light of the latest regulatory environment?

Thank you.

管理层晚上好,感谢接受我的提问。我这个问题是关于云收入展望方面的。首先非常感谢管理层提供的指引,即在2030年实现云收入1,000亿。但是考虑到最新的BIS宣布的事项,我们应当如何看待本财年接下来两个季度云收入的增长路径?然后另外我们是不是有信心可以在下一个财年实现超过50%的收入增长,即FY28?那么如果是有这样的信心,考虑到现在的情况,这个信心是来自哪里?有什么依据?谢谢。 好,谢谢您的问题。在我们2030年的一个长期目标的整个规划当中,我们的AI基础设施其实一直都是基于我们平头哥的国产芯片的供应链计划。所以随着产能的爬坡,国产算力在我们的整体算力中的比例将会持续提升,将会成为支撑云收入增长的主要引擎。当前的AI算力在全球仍是供不应求,尤其是在中国,我们觉得近两年会尤其的紧缺。那随着以平头哥为核心的AI全栈产品及供应链的持续推进和放量,我们认为之前提到的2030日历年云外部收入1,000亿美元和20%以上的利润率的目标,还是有很强的可见性。

Thank you. Our long-term business plan through to 2030 has always been built on T-Head's chip supply road map.

As T-Head's capacity ramps up, domestic compute will contribute a higher proportion of our overall compute and will become the primary engine supporting cloud revenue growth, subsequently lowering dependence on overseas compute. The market does remain supply constrained.

That's true globally and especially here in China, I think, over the next couple of years. As we continue to advance our full-stack AI products centered on T-Head and supply chain, we see strong visibility towards the previously communicated calendar year 2030 targets of $100 billion in external cloud revenue and EBITA margin of 20%

Eddie Wu

plus. 好,至于说到,就是明年的这个情况,我们认为现在全球的算力出口管制的新规其实影响明年的产品。那全球的整个供应链我觉得有非常大的不确定性。但是另外我们在国产算力上的进度的推进还是超出我们原来的预期。所以综合各个情况来看,对于明年的经营规划目标可能会产生一些影响,但我们还在评估当中。但综合各种因素,我们预期下一个财年我们的经营目标范围会落在30%到45%的增长范围区间。同时随着2028日,2028的日历年,我们的国产算力产品规模的一个持续的放量扩张,预期在2028日历年,我们会重新恢复50%以上的高速增长。

Turning to next year's outlook, the new compute export control rules will affect next year's products and services, and there is still uncertainty as regards the supply chain. However, our own progress in developing domestic computing power is, in fact, already ahead of our original expectations.

Taking all these different factors into account, our operating plan and targets for next year may be affected to some extent, but we're still in the process of evaluation right now. Taking all these factors into account, we might expect our growth target for cloud revenue to be in the 30% to 45% range.

As the scale of our domestic compute products expands in calendar year 2028, we expect cloud revenue growth to rebound to 50% or higher in calendar year 2028.

Gary Yu

Thank you.

Lydia Liu

Next question, please.

Operator

Your next question comes from Kenneth Fong with UBS.

Kenneth Fong

Hi. Good evening, management, and thanks for taking my question.

I actually have a follow-up also regarding the tighter US restriction affecting the access to advanced GPU and overseas compute. How should we think about the capex trajectory to reach our compute target of 20 gigawatts?

Understand the situation is still very dynamic with the potential change in our capex cadence, but would it be possible that we will turn free cash flow positive earlier than expected? Thank you.

管理层晚上好,感谢接受我的提问。其实我还是想接着刚才的话题进一步地来展开问一下,就是鉴于现在美国对先进GPU实施限制,包括海外的算力资产实施限制,那我们应当如何看待我们的资本开支路线图,就是将来如何来实现我们20吉瓦的一个算力目标?当然现在的情况还是不停地在变化,而我们资本开支的速度有可能也会相应发生变化。但是有没有可能我们会比原来预期的时间更早能够实现正的现金流呢?谢谢。 好,谢谢你的问题啊。对,我觉得现在的全球的各个区域的整体的算力政策确实有非常大的一些变化。对于整体我们长期的长远阿里云的国内已经全球的数据中心的建设肯定会产生一些影响。那就我们现在的初步评估来看,对于长远20吉瓦的目标,应该是有可能会降低这个目标。所以同时对于我们整体的capex计划,应该也会是有一定的,有一定的降低。但是我们会在国产算力上,国产算力相关的供应链上投入更积极的,更积极的推进的这个整体的计划。所以在现在我们初步的评估计划当中,如果说我们预期在随着2027、2028日历年的平头哥的为基础的这个算力供应链的持续的放量和放量的节奏,我们预期的目标是在2028年的下半年能够实现现金流的转正,也就是一个相当于我们最大的一个持续的投资周期会在2027年和2028年的上半年。但是随着2028年之后我们彻底实现就是在国产算力供应链的大规模的持续扩量,那我觉得我们会经过AI基础设施的一个投资,投资最大的这一个阶段。所以对于我们长期的现金流规划,我们觉得还是,还是非常有,有这个信心,对。

Thank you for your question. Indeed, I believe that global policies around compute across different regions are undergoing significant changes at present.

From a long-term perspective, certainly the construction of data centers will experience some impact. Based on our current preliminary assessment, it may lead to a downward revision of that long-term 20-gigawatt target.

To that extent, there will also be a certain impact on our overall capex plan. Despite that potential reduction, we will be more actively advancing our overall plan, particularly in terms of developing domestically produced compute power and the related supply chain.

Our current preliminary assessment would be, we could say our expectation would be that in calendar years 2027, 2028, we will be investing most aggressively in the ramping up of our domestic chips and that we should be able to achieve positive cash flow overall by 2028 in the second half of that year. Basically the most sustained investment cycle will span the first half of 2027 and 2028, but once we get there and achieve large-scale continuous output of domestically produced chips, then that will lead to an easing of that level of spending.

Yes, there will be some long-term impact on our capital expenditure plan.

Kenneth Fong

Thank you.

Lydia Liu

Next question, please.

Operator

Your next question comes from Joyce Ju with Bank of America.

Joyce Ju

Good evening, management. Thanks for taking my questions, and congrats on the very solid quarter.

My question is on your core e-commerce business. Given the soft macro environment and a cautious peer guidance for the fourth quarter, what is your December quarter outlook for Taobao Tmall in terms of CMR growth and adjusted EBITA trajectory?

On instant commerce, could you update us on your progress in the December quarter and your visibility on the timeline for further loss reduction next year? Thank you.

感谢。我的问题呢是关于核心电商业务的。那么也祝贺你们取得了本季度非常扎实的业绩。那么我的问题是这样,鉴于当前宏观环境出现疲软,以及你们对于第四季度给出的是偏谨慎的这样的指引,那么对于12月份季度淘天集团的整体的展望如何,尤其是CMR增长以及经调整EBITA的这个发展路径?然后关于即时零售方面,是不是可以给我们提供12月份季度最新的一个进展的介绍,以及你们对于明年进一步减亏的一个预期的时刻表?谢谢。 好,我来回答这个问题。对,首先三季度我们整体电商板块的表现是符合预期的。国内电商保持稳定,其实即时零售板块在大幅减亏的情况下依然保持着非常好的同比。其他业务板块的盈利水平也明显改善。刚才你也讲到,就是我们也看到整个四季度,包括三季度、四季度以来,整个国内的消费大环境还是依然承压。消费者的支出意愿趋于保守。同时部分品类我们也一直面临着比较激烈的市场竞争。所以四季度以来,平台的主要经营指标表现与中国实物商品网上零售的趋势一致。在当前大消费环境下,刚需品类较可选消费品类展现出更好的趋势。平台也将进一步强化刚需品类在日常销售的竞争力。

Thank you. In the September quarter, our e-commerce business performed in line with expectations.

Domestic e-commerce remained stable. Quick commerce grew well on a sequential basis while narrowing losses substantially.

Profitability across our other businesses improved meaningfully. As you pointed out in your question, yes, indeed, the domestic consumption environment has remained under pressure in the December quarter.

Consumers are becoming more cautious in their spending, and certain categories continue to face intense competition. Quarter to date, our key operating metrics have been in line with the trend of national online physical goods retail sales.

In the current consumption environment, consumer staples are showing healthier growth than discretionary products, and we will continue to further strengthen our competitiveness in staples during regular selling periods.

Jiang Fan

然后即时零售依然是未来几年这个平台增长的核心驱动力。那三季度即时零售非餐饮部分线上交易额同比增长40%,领跑行业,尤其是前置仓业务发展迅速。在食品生鲜、日用百货等品类上,即时零售持续改变用户的消费习惯,带动相关品类。餐饮方面,我们的UE优化符合预期,市场份额保持稳定,客单价持续上涨。

Turning to quick commerce, quick commerce will remain a key driver of platform growth over the next few years. In the September quarter, GTV of online non-food categories of quick commerce grew more than 40% year-over-year, outpacing the industry.

In particular, our self-operated front warehouses expanded fast. In categories such as food and fresh produce and daily necessities, quick commerce continues to reshape consumer habits and drive growth in these categories.

On the food delivery side, our unit economics improved in line with expectations. Market share remained stable, and average order value continued to rise.

Jiang Fan

12月季度我们预计即时零售的亏损会持续缩小。然后我们继续维持明年我们的亏损缩小一半的这样的一个目标,并且于29财年实现即时零售的盈利。

In the December quarter, we expect our losses in the quick commerce business to continue narrowing. We remain committed to our goal of halving losses next year and achieving profitability in instant retail by fiscal year 2029.

Jiang Fan

然后在B2C电商跟即时零售之外,我们电商板块中的其他业务,例如B2B、闲鱼等业务,近两年来也持续进步。对,本季度我们整合了这个B2B的国际与国内业务板块,打通国内供给海外需求。在AI技术的持续加持下,我们相信此整合也会进一步释放B2B板块更大的潜力。闲鱼也在上个季度实现了29%的订单同比,并且实现多季度盈利。那我们也相信这些业务将来会为整体电商板块带来持续的营收增长跟利润贡献。

Beyond B2C e-commerce and quick commerce, other businesses within our e-commerce segment, including B2B and Xianyu, have also made solid progress in recent years. This quarter, we integrated our international and domestic B2B businesses to better serve overseas demand with domestic supply.

By leveraging AI, we believe this integration will unlock greater potential across the entire B2B business. In the September quarter, Xianyu delivered 29% order growth and achieved continued profitability.

We believe these businesses will continue to contribute incremental revenue and profit to the overall e-commerce segment going forward.

Lydia Liu

Thank you. Next question, please.

Operator

Your next question comes from Robin Zhu with Bernstein.

Robin Zhu

Hi. Thanks, management, for taking my question.

I guess there's two slight parts to my question. One is, could management discuss the impact of the BIS restrictions on model training and therefore the speed of iteration of the flagship Qwen models?

Then related is, I guess our understanding is that a proportion of your data center capacity expansion is going to be funded by partners, T-Head customers. Could you share an update on progress towards deals of this nature, potentially as a proof of concept?

Thank you.

感谢管理层接受我的提问。我这个提问分两个部分。第一,就是是不是可以请你们讨论一下BIS的新限制如何影响模型的训练,会不会由此影响到我们旗舰通义千问模型迭代的速度?其次,与之相关的,就是据我们了解,你们扩大数据中心的这个能力,一部分的资金会是来自于你们的合作伙伴,也就是你们的客户。那是不是可以分享这一方面的一些最新的进展,是不是跟一些客户开始达成这样的一些交易,可以来验证这样的概念?谢谢。 谢谢您的问题啊。就是,我觉得新的,新的算力获取的一个规定正在落地当中,很节,很多细则。那我们现在的情况,我们是在各地,在全国各地使用算力,或者说是一贯遵守适用的法律法规,所以对我们的产品以及对财务的具体的影响,可能还需要一些具,具体的信息才能够一个比较完整的评估。但是如果影响到后面的一些产品的进度的话,我觉得也会是一个比较短期的影响。因为在2027年的,在2027年的一季度,其实我们自己的,我们自己的基于平头哥的国产芯片,这个国产芯片是完全可以支持大规模的,超大模型的一个训练。我们现在其实已经在开始在我们现在的这一代的,就今年这一代的平头哥芯片上开始去训练下一代的大型模型。那我觉得随着我们平头哥的国产芯片的产能的持续扩量,我觉得我们从长远来看,可能对我们的模型的发展其实是一种,就是会,会更是一种长期的支撑吧。因为未来的话,我们觉得大部,我们的大部分,包括我们自己,包括我们的合作伙伴,包括我们的客户,绝大部分的模型都会完全依赖于国产,国产芯片去做整个的训练,无论是预训练还是,还是后训练。所以随着我们2027、2028年的产能的持续扩量,我觉得对于我们模型的长期持续迭代,这个节奏反而更有,反而更有可控性。

Thank you for the question. We're in a stage now when new regulations regarding compute power and the acquisition of compute power are being implemented and with numerous detailed rules that will be coming out.

What I can say as regards Alibaba's situation is that we have obtained and used compute in compliance with applicable laws and regulations, and of course, will continue to do so. Because these are new rules coming out now, it's too early at this stage for us to evaluate the impact.

As I was saying earlier, we expect to ramp up very significantly production capacity of T-Head chips in the slightly longer term, certainly by 2027. Our current T-Head chips are already capable of fully supporting the training of large-scale, including ultra-large-scale models.

In fact, this year, we're using this year's chips already to be training the next generation of models. As T-Head production capacity ramps up, we expect that will be providing the long-term support for model development for both us, for our partners, for our customers.

It'll be domestic chips that will be underpinning all of that training, including pre-training and including post-training. That is when we expect to see that ramp up in 2027, 2028, and I think once we get there, we'll actually be in a better position to control and pace the speed at which we iterate our foundation models.

Lydia Liu

Okay. Next question, please.

Operator

Your next question comes from Ellie Jiang with Macquarie.

Ellie Jiang

Right. Thank you so much for taking my question.

I have a question mostly regarding T-Head. We're very encouraged to hear management commenting on the solid ramp-up pace for T-Head and obviously the promising pipeline.

I'm just wondering what proportion of the incremental training and inference capacity can T-Head supply over the next few years? Also, just a quick follow-up, can management comment on the latest updates regarding T-Head spinoff and IPO timeline?

Once this sort of comes out, how do we really assess the internal versus external resource allocation for T-Head, and what would be the implication on cloud margins? Thank you.

好,感谢管理层接受我的提问。我这个问题是想更多地了解一下刚才说的平头哥的这个问题。我们听到平头哥这个国产芯片生产产能在这个不断地扩大,这个是令人感到鼓舞的,也有非常棒的未来的这个产品的管线。那我想知道就是接下来几年,就是我们如果是说到增量的训练、增量的这个数据输入,它多大的一个百分比会是给到通义千问的?另外一个后续的提问,就是管理层是不是可以介绍一下这个平头哥分拆上市IPO的一个进展。那么当有关的细节公布以后,我们应该如何正确地评估平头哥这种内部外部的资源的配置,然后对于云业务的利润率会带来什么样的一个影响?谢谢。 好,谢谢你的问题啊。就是,确实我们觉得从,就现在这个阶段,新的情况来看,我觉得我们在2027以及2028年,在2027之后,其实我觉得我们绝大部,大部分的新增的这个,新增的AI芯片的这个供给应该都是以我们自己的国产的平头哥的芯片为主,然后可能我们也会,也会采购一些其他的,其他的国产芯片来支持我们,我们自己内部的这些推理需求。所以从这个角度上来看,我们会,我们自己的旗舰的芯片会优先满足于大模型的,大模型的训练,然后我们在,包括我们自己,也包括我们的重要的深度的合作伙伴。对,所以我们现,会也会是一个相对平衡的一个分配逻辑,就是我们会首先优先保障我们自己的模型的迭代,然后在模型的迭代基础上,我们也会考虑我们重要的合作伙伴的客户的需求。同时在推理层面上,我们会尽量地去采购更多不同构成的这个,更多不同构成的这个芯片,也包括很多其他品牌的国产芯片。所以这是我们大概的一个,一个预期计划。

Thank you for your question. Looking ahead to 2027, 2028, I think after 2027, most of the incremental AI chip capacity we will add will be primarily domestic, our proprietary domestic chips.

At the same time, we will also purchase other chips from other domestic suppliers primarily to support inference workloads. Our own flagship proprietary chips will be used as a first priority to support the training and iteration of our foundation models, but also to support training for our close partners.

It'll be a balanced allocation where the first priority, of course, will be to support our own model training needs, and as long as our own model training needs are satisfied, then of course, to provide resources to our important customers. At the same time, we will also be supporting inference workloads with a wider variety of domestic chips, including domestic chips from other brands.

Eddie Wu

好,关于您说的对云业务的利润率的这个影响。我觉得分,分几个层面吧。第一,我觉得未来在2027和2028年,在中国,中国国内的这个,这个市场的话,算力可能是最紧缺的这两年吧。这两年当中,所以我们觉得由于算力的这个紧缺,总体来说,对于总体算力的价格毛利上面其实是相对来说我觉得是可以保障的。那另外一方面,由于总体来说都是我们自研的芯片,再加上我们其实规模的持续放量,所以我们在采购和供应链方面其实相对来说也还是有一个不错的优势。所以我们比外采的芯片来做云业务和AI业务,肯定的利润率相对会是一个比较正向的,比较正向的一个提升。

The other part of your question regarding profitability of the cloud business, I think that period 2027, 2028 will be the period in which compute supply is at its scarcest in China's domestic market. I think that is actually very favorable when it comes to overall compute pricing and gross margin for our compute offerings.

At the same time, because we are using our own proprietary domestic chips and ramping up very significantly production capacity, there will be cost advantages there and across the supply chain in general. I think these factors are favorable when it comes to thinking about gross margin.

Eddie Wu

对,对于您关于这个平头哥分拆IPO的这个问题我来回答一下。这个我们其实不久前刚刚做了公告,那平头哥现在是一个保密的形式向相关的监管机构提交了上市申请。那现在是基于上市合规的要求,我们在目前这个阶段是没有办法披露太多的这个平头哥相关的财务信息,包括其他未公开的信息。但是可以确认的是,国产算力在阿里云整体算力中的占比会持续提升。那这一趋势对公司整个云业务的结构性是肯定是利好的。

On your question about T-Head spinoff and IPO, we recently announced that T-Head has on a confidential basis submitted a listing application to the regulators. Given listing compliance requirements, we are unable to disclose T-Head's financial information or other non-public information at this stage.

What we can confirm is that domestic compute continues to account for a growing share of Alibaba Cloud's overall compute capacity, and that's a trend that is structurally positive for our cloud business.

Lydia Liu

Next question, please.

Operator

Your next question comes from Ronald Keung with Goldman Sachs.

Ronald Keung

Thank you, Eddie, Toby, Fan, and Lydia. Pleasant to see our sequentially narrowed loss for the AI Labs and Application segment.

Just want to ask about, we're seeing other China internet giants are actually ramping up on the investment scale on the model and application layer. How does management see the AI value eventually?

Would it kind of eventually be at the application agent layer, and therefore the investment priority for Qwen Work? How do we differentiate Qwen Work and the Qwen models, and the implications for this segment's investment scale loss for the next few quarters and one or two years?

Thank you.

非常感谢管理层接受我的提问。我们就是很开心注意到我们AI实验式应用方面这个亏损是环比收窄。然后同时我们看到很多其他中国大型的互联网公司,他们其实是在扩大规模去投资,尤其在模型和应用层。那我想知道管理层如何看待最终AI的价值会主要沉淀在哪一层?觉得主要会不会是在应用智能体这一层呢?那么如果是,我们应该如何来配置在通义千问方面的这些投资?然后就是通义千问和其他的通义模型的差异化,我们应该怎么去了解?然后这种差异会如何影响投资的规模以及这个分布,接下来几个季度或者是未来一两年的这个亏损的轨迹?谢谢。 谢谢您这个问题啊。其实这个问题其实我觉得现在行业内,大家不同人也有不同的看法,就是包括不同的公司也有不同的看法。我觉得就现在我们觉得还是AI的早期阶段。那现在这个应用和模型到底未来是哪一个更重要,其实我觉得,其实可能这两者需要结合在一起看。因为我觉得在未来的模型和应用一定需要联合发展,或者说是他们是一个共同的协同发展。模型去帮助应用,或者应用去能够协同模型的一个增长。但是如果一个应用,如果它没有一个更强的模型,或者一个独特的更强的模型和它联合去做,我觉得这个应用长期的用户留存其实还是存疑的。就像说如果一个应用背后大家用的都是相同的模型,那我觉得这个应用的可替代性也会非常强。所以我觉得就我们的看法来说,我觉得未来还是需要模型和应用联合发展,能够去创造一个更独特的,产品价值,或者更独特的一个产品力。这样的话才能够长期地留存在企业内部,或者说客户的桌面上。所以现在来说,我们自己的策略是强调我们的应用和模型联合发展,怎么样提升更强的这个产品能力。所以现在比较激进的用户推广,或者说比较激进的用户补贴上面,我们现在还是相对比较克制。长期来看,我们更看好一个独特的模型和一个独特的应用,创造一个独特的对客户的一个价值,是一个比较难以被替代的一个价值。这样的应用和模型的联合体才能够是一个比较有用的长期价值。

Thanks for that question. I think it's fair to say that within the industry, different people have different views on that.

Different companies will have different views of that, and at this point, I think we're very much still in the early stages of AI development. When it comes to the relative importance of apps versus models, which is more important, I think you need to look at both together, and I would say that apps and models need to be developed in synergy, one with the other, and that's our approach.

Because if an app isn't built on a stronger, better, more unique model, if all the apps are basically running on the same model, then those apps are very easy to replicate, and it's difficult to establish a long-term unique value proposition for that app to retain users and keep that app as the app being used in the enterprise or being used on someone's desktop. We stress the joint development of apps and models and thereby aim to achieve differentiated product capabilities based on their synergies.

We've been very restrained when it comes to promotion and subsidies, and we really place the emphasis on delivering unique value that's difficult to be supplanted or replaced or caught up to by a competitor. We think the best way to do that is by developing models and apps in tandem and in synergy.

Thank you.

Lydia Liu

Next question, please.

Operator

Your next question comes from Thomas Chong with Jefferies.

Thomas Chong

Hi. Good evening.

Thanks, management, for taking my questions. My question first is about our ARR trend.

Given that we have highlighted that ARR exceeded a new milestone, RMB20 billion in August, just want to get some more color with regard to our full-year target. Is it going to beat the 4.5 billion or RMB30 billion by the end of the fiscal year?

On that front, can management comments about the recent trend in ARR and our thoughts about the competitive landscape on fresh model getting increasingly popular? My second question is about our overseas strategy.

Can management comments about our thoughts about the cloud revenue mix between domestic and overseas in coming years? Thank you.

感谢管理层给我提问的机会。我这个问题呢是关于我们年化收入ARR的趋势。因为我们前面凸显了一个最新的成绩,就是8月8月份这个ARR是突破200亿人民币。那么我想做更具体的了解,就是是不是可以介绍一下?然后就全年的目标而言,你们是不是认为这个收入可以财年底的时候,这个ARR是不是可以达到300亿人民币?那么这一方面相信管理层在密切关注ARPU的趋势,并且在分析竞争格局,那是不是可以介绍一下当前的一个竞争格局呢?第二个问题就是关于海外我们的战略,是不是可以介绍一下云业务方面这个国内国外的收入占比的展望? 好,谢谢您的问题。就是总体MaaS业务的收入还是比较依赖于新模型的,新模型的推出。那现在其实最在国内总体最近这一两个月,其实包括我们自己,包括我们的,我们的合作伙伴,其实总体新模型的推出的进度,其实相对来说,其实是有些就不如我们原来的预期的。那我们自己的千问4也刚刚发布,同时我们自己的下一代的视频模型,我们对下一代的视频模型也寄予厚望,也会在近期开始发布。所以随着我们的新模型以及包括我们合作伙伴的新模型的逐渐发布,我觉得我们ARR的一个,一个增长会,会很快开始就有,有一个比较,比较快的增长。我们在预期在12月份,所以我们在预期全,整个财年的300亿的ARR的目标还是,还是没有改变。

Thank you. When it comes to MaaS revenue, I think this does tend to a great extent to rely on new model releases.

Over the past two months or so, I would say that new model releases from us and from our partners have lagged a bit behind our own original expectation, but we've just recently launched our new flagship Qwen 4 model, and our next-generation video model is just around the corner and for which we have extremely high hopes. I think that we are very confident in achieving our target of $3 billion in ARR for the full fiscal year.

That target remains unchanged. 30, RMB30 billion, RMB30 billion.

Thank you.

Eddie Wu

好,关于您说的那个海外云的业务的一个影响。其实我们现在也正在评估算力出口规则对海外云业务的一个影响。那所以我们觉得可能还是具体的情况还需要一些时间去分析和评估。但是目前阿里云在全球32个区域运行的109个可用区,所以包括我们在荷兰的数据中心,包括我们在法国和巴西的数据中心都在逐渐的,逐渐的开服。那在未来的,我觉得在未来的海外的阿里云的海外业务继续将会是一个我们持续的重点。但是在未来海,海外业务当中,AI的部分其实我们会有各种可能不同的合作方式,也许就是需要更多的海外的合作伙伴,或者是甚至长远来看,我觉得我们自己的芯片去做出口,也是我们一个未来可行的一些方案。当然这些都可能是要在满足国内的需求之后的這個基础之上。所以但是我们长远来看,我们对于像阿里云这样的全栈的,全栈的基础设施,首先服务于我们的总体的中国公司出海,以及以及在海外的这些客户,我觉得我们对这个业务的前景还是长期保持一个比较强的一个信心和一个业务发展的一个期待。 Then

turning to your question about overseas cloud business, we're still currently assessing how new regulations on compute exports will affect that business, and it will take a bit of time to analyze and assess those impacts. At present, Alibaba operates 109 availability zones across 32 regions worldwide.

This includes data centers in the Netherlands, France, and Brazil that we're progressively bringing online. Looking to the future, I think that developing the overseas business for Alibaba Cloud will be a sustained focus going forward.

That could involve a number of different partnership models. We can work with different overseas partners in different ways.

I believe that down the road, exporting our own proprietary chips is also a viable option for us. Of course, we would only pursue that kind of initiative after having fully satisfied domestic demand.

Looking ahead, we view Alibaba Cloud as a comprehensive full-stack infrastructure, and it can support Chinese companies' global expansion as well as overseas customers. We continue to have strong confidence in the long-term prospects of this business, including overseas, and look forward to its continued growth.

Lydia Liu

Operator, let's take the last question.

Operator

Your next question comes from Alicia Yap with Citigroup.

Alicia Yap

Hi. Good evening, management.

Thanks for taking my questions. I wanted to follow up on the cloud margin.

Can management elaborate the key gross drivers with the decent improvement of the cloud margin this quarter? How should we be thinking about the cloud margin trends in the coming quarter?

Also given the latest constraint on the cloud compute environment, is there a chance that Alibaba might able to increase the cloud pricings to meet the stronger demand? Then would there be any chance on the AI lab losses could also be narrow at a faster pace amid the latest change of the compute environment?

Thank you.

晚上好,感谢接受我的提问。我想就云业务利润率这一方面在后续提问。首先想了解一下就是本季度这个云利润率提升背后的主要因素有哪些?关键动能有哪些?以及接下来几个季度这个利润率是不是会有进一步的这种提升?然后就是考虑到现在这个算力的这个管制加强、加紧,是不是可以使我们有机会可以进一步地提高云产品、云业务的一个定价?然后与之相关的是,我们这个AI实验室这一方面的这个亏损是不是有望可以更快地来收窄? 好,感谢您的问题啊。就是我们现在来看,我觉得有几个因素驱动了AI云业务的利润率的一个持续的改善。那我觉得一个主要的一个因素就高利润率的AI相关产品就占我们的整体的一个经常性收入的这个比例其实是在持续的提升。那这样的话其实我们相对的来说就整体改善了我们的云业务的,云业务的整体利润率。当然另外一个方面,由于全球的算力紧缺,以及刚才我已经说到了,尤其在中国更,更大的一个算力紧缺的行业,行业结构下,我们其实各方面的产品的价格,价格的和客户的这个谈判筹码,也就是相对来说我们对,有一个比较好的一个毛利空间。所以每个产品基本上都会有一个对比去年来说都会有一个比较好的毛利空间。那另外一方面,就是随着我们的,随着我们整体的数据中心的扩张,我们带来的整体的规模效应也在持续的,持续的扩张,无论是我们的供应链的规模,还是说我们在客户这边,因为客户用了我们多个产品之后,它的迁移成本会越来越高,客户这边的一个持续的一个接受程度,这些都在持续的改善我们的利润率的一个,都在持续地改善我们一个利润率水平。那我觉得从更长期来看,我觉得我们有一个非常,由于AI的一个,AI的一个整体的一个技术的发展,那我觉得AI云,尤其是带自研,自研芯片的一个AI云业务,我觉得和我们原来传统的通用计算的云业务产生了一个结构性的改变。因为当有了自研芯片之后,其实尤其是再有自己的模型,在有一个全栈的AI服务之后,其实阿里云的业务其实已经变成了一个非标准化的产品。对于客户来说,其实它很难在某种程度上进行完全的像一个标准化产品一样去进行比价,或者说对它终归是有一些不同的,对于阿里云全栈的技术来说,技术来说它有一些不同的需求,所以它很难。我们最近也见了不少客户啊,就是客户我觉得AI时代到来之后,阿里云对他来说不是一个供应链的采购,采购商,阿里云对他来说是一个共同开发产品的合作伙伴。我觉得这是一个对于长期来说,AI对于我们的阿里云的业务来说是一个结构性的改变。那这样的话,其实我们对于客户的全方位的产品支持,同时客户对我们的长期的接受度,这些都会以及其实因为产品的差异化,其实也很难进行直接的招标和比价。那我觉得这个是一个对我们长期的AI云业务的一个利润率的一个结构性的改善。

Thank you for the question. I think there are several different factors that have contributed to the margin expansion in our cloud and AI business.

I think the first of these is that the proportion of high-margin offerings within our recurring revenue has increased. Overall, that is clearly driving improved cloud margins.

We also talked earlier about the global shortage of compute that is affecting the entire industry everywhere, and I think, again, that does give us a certain pricing advantage. It gives us a It puts us in a stronger bargaining position when it comes to pricing with customers, which, again, has contributed to better gross margin than the last year, for example.

I think an additional factor is that as we continue to build out our our data center footprint, we achieve greater economies of scale, not just in terms of our own resources, but also in terms of the number of different products for which customers rely on us. As a customer uses more of our different products, the cost of switching becomes higher for that customer.

Again, this is also, and these factors together are structurally positive for our cloud and AI margin. I think looking ahead to the even even longer term, as we continue to develop AI-related technology and upgrade our AI-plus-cloud business, building on our own proprietary chips, that, again, is structurally advantageous for our gross gross margin.

In particular, with our AI offerings running on our proprietary cloud on our own proprietary chips, it's really a non-standard or non-commodity product. Alibaba is really becoming more like a partner with many of our customers, co-developing products for them, it's not a commoditized offering that customers can easily just compare against the competitors on price alone.

These are structural changes that will further serve to differentiate our product offering and make it more difficult, I think, for competitors to to catch up.

Lydia Liu

All right. Thank you for joining us today, and we look forward to speaking with all of you again soon.

Operator

That does conclude our conference for today. Thank you for participating.

You may now disconnect.